v1.2.0 - Multi-Provider Vision API Support

✅ Added
OpenAI-compatible API provider adapter - Support for llama.cpp, vLLM, LocalAI, Jan, and any OpenAI-compatible vision API (#8)
Provider abstraction layer - Clean architecture with VisionProvider base class
OllamaProvider - Default provider (backward compatible, zero config changes required)
OpenAICompatProvider - Handles OpenAI /v1/chat/completions format with automatic payload translation
Environment variable configuration - FIXXER_API_PROVIDER and FIXXER_API_ENDPOINT for runtime provider selection
Config file API section - [api] section in ~/.fixxer.conf for provider settings
Configuration priority system - Environment Variables > Config File > Defaults
🔄 Changed
Vision API calls now use provider abstraction instead of direct Ollama integration
Connection checking is now provider-agnostic
Default behavior unchanged - Ollama remains the zero-config default
Fixed
Improved error handling for vision API failures across different providers
✍️ Technical Details
New vision_providers/ module with factory pattern for provider instantiation
Payload translation from Ollama format to OpenAI multi-modal content structure
Graceful fallback for connection checks (tries /v1/models then base URL)
Maintains full backward compatibility with existing Ollama workflows

FIXXER v1.1.0 - Modular Architecture Refactor

FIXXER v1.1.0 - Modular Architecture Refactor

🏗️ Major Refactor

Split monolithic engine.py into focused, maintainable modules:

• config.py - Configuration management and app settings
• security.py - SHA256 hash verification & sidecar file generation
• vision.py - AI/Ollama integration & RAW processing (rawpy)
• engine.py - Workflow orchestration (now focused & clean)

🔧 Improvements

• Cross-platform defaults - Default destination now ~/Pictures/FIXXER_Output (works on macOS/Linux/Windows)
• Case-insensitive file matching - Finds both .jpg and .JPG files
• Removed deprecated tools - Replaced dcraw with rawpy (Python-native, cross-platform)
• Cleaned up dead code - Removed MockSessionTracker, check_dcraw(), etc.
• Fixed Easy Button workflow - Simple Sort now works correctly

📚 Documentation

• Updated README with modular project structure
• Clarified system requirements (rawpy/exifread via pip)
• Cleaned up .gitignore duplicates
• Version metadata synced across all files

🎯 Type

Architecture refactor + documentation sync - backward compatible

---

Built with precision. Secured with cryptography. Powered by AI.

FIXXER ✞ v1.0.1 - SRC Layout & Homebrew Support

What's New

• Migrated to professional src/ package layout for reliable pip/Homebrew installation
• Added Homebrew tap support: brew tap bandwagonvibes/fixxer
• Improved CSS asset loading via importlib.resources
• Added python -m fixxer module entry point

Install

   brew tap bandwagonvibes/fixxer
   brew install fixxer

Or via pip:

   pip install git+https://github.com/BandwagonVibes/fixxer.git

FIXXER ✞ v1.0 - "CHAOS PATCHED // LOGIC INJECTED

🎯 FIXXER v1.0 - Professional Photography Workflow Automation

Evolved from photosort to FIXXER ✞ PRO

A major upgrade bringing professional-grade features, cryptographic integrity verification, and dual UI modes for serious photographers who demand both speed and safety in their digital asset management.

---

✨ What's New in v1.0

🔐 Hash-Verified File Operations

• SHA256 integrity checking on every file move operation
• Halt-on-mismatch protection prevents silent corruption
• JSON sidecar files (.fixxer.json) create permanent audit trails
• Production-tested: 120+ files, zero corruption events

🎨 Pro Mode - "Phantom Redline"

• Toggle with F12 for tactical precision dashboard
• Milestone HUD with real-time workflow stats (BURSTS, TIER A/B/C, HEROES, ARCHIVED, TIME)
• Redline sparklines for system monitoring (RAM/CPU)
• Minimal, professional aesthetic for focused work

📷 120+ RAW Format Support

• Python-native RAW processing via rawpy (no external dependencies)
• Cross-platform: Linux, macOS, Windows
• Supports: RW2, CR2, CR3, NEF, ARW, DNG, RAF, ORF, PEF, SRW, and 100+ more
• Zero temp files - pure in-memory processing

🤖 AI Vision Integration

• Ollama-powered naming (qwen2.5vl, llava, bakllava)
• Semantic burst detection using CLIP embeddings
• AI session naming from visual analysis
• Creative critique mode for artistic feedback

📊 Quality Analysis Pipeline

• BRISQUE quality scoring for sharpness assessment
• Exposure analysis (crushed blacks, blown highlights)
• Automated culling into Tier A/B/C folders
• Smart burst grouping based on visual similarity, not timestamps

---

🔧 Bug Fixes & Improvements

Fixed:

• ✅ Unicode corruption in photosort_engine.py (emojis/symbols now render correctly)
• ✅ Pro Mode logo display in TUI
• ✅ Cross-platform RAW file handling (replaced macOS-only sips)
• ✅ Thread safety in logging and progress updates

Improved:

• ✅ Professional README with architecture diagrams
• ✅ Modern Python packaging (pyproject.toml)
• ✅ Comprehensive dependency documentation
• ✅ Smart .gitignore (protects user photos and config)

---

📦 Installation

Requirements:

• Python 3.10+
• Ollama (for AI vision features)

Quick Start:

# Clone the repository
git clone https://github.com/BandwagonVibes/fixxer.git
cd fixxer

# Create virtual environment
python3 -m venv venv
source venv/bin/activate  # On Windows: venv\Scripts\activate

# Install dependencies
pip install -r requirements.txt

# Pull an AI vision model
ollama pull qwen2.5vl:3b

# Launch FIXXER
python3 photosort_tui_styled.py

---

📚 Documentation

• README.md - Complete feature guide
• requirements.txt - Dependencies with platform notes
• Configuration - Stored in ~/.photosort.conf

---

🎯 Workflows Included

Workflow	Description
Auto	Complete pipeline: burst stack → cull → AI name → archive
Bursts	CLIP-based semantic grouping + AI naming of best picks
Cull	Quality analysis → Tier A/B/C sorting
Stats	EXIF insights (cameras, lenses, lighting conditions)
Critique	AI creative feedback (composition, mood, suggestions)
Easy Archive	Simple AI naming + keyword folders (no culling)

---

🔒 Security & Integrity

Every file operation is protected by:

• SHA256 hash calculation before move
• SHA256 hash verification after move
• Automatic sidecar file generation (.fixxer.json)
• Workflow halts on corruption detection (RuntimeError)

Example sidecar file:

{
  "fixxer_version": "1.0",
  "filename": "golden-hour-cityscape.jpg",
  "sha256_source": "a1b2c3d4...",
  "verified": true,
  "timestamp": "2024-11-20T14:35:22.123456"
}

---

🎨 UI Modes

Toggle between two aesthetics with F12:

Standard Mode (Warez):

• Red/white/black color scheme
• ASCII art branding
• Motivational progress phrases
• High-contrast buttons

Pro Mode (Phantom Redline):

• Tactical black aesthetic
• Clean typography
• Real-time HUD dashboard
• Minimal, professional design

---

🙏 Acknowledgments

Built with:

• Ollama - Local LLM inference
• rawpy / LibRaw - RAW processing
• Textual - Modern TUI framework
• OpenAI CLIP - Semantic burst detection
• BRISQUE - Image quality assessment

---

📄 License

MIT License - See LICENSE for details.

---

🚀 What's Next

Future roadmap considerations:

• Native Windows support (non-WSL)
• GPU acceleration for CLIP embeddings
• Lightroom Classic plugin integration
• Additional AI vision models
• Batch processing optimizations

---

Built with precision. Secured with cryptography. Powered by AI.

✞ FIXXER PRO - From chaos to order, one hash at a time.